[
https://issues.apache.org/jira/browse/HIVE-8561?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14200562#comment-14200562
]
Brock Noland commented on HIVE-8561:
------------------------------------
Hi [~nyang],
Thank you for the annotations! Since the CBO is such a huge component and in
it's infancy, I feel like {{Unstable}} might be more appropriate than
{{Evolving}}. However, before making that change I think we should settle with
[~jpullokkaran] the correct way to perform this integration.
bq. Why can't Drill be plugged in as another execution engine just like MR,
TEZ, Spark?
[~jpullokkaran] It's reasonable for Drill to add API's in order to use the
query plan. The Drill project like many other projects are users of Hive. As
mentioned previously, it's important to agree upon some kind of api visibility
and stability. Na has agreed to an unstable interface ("It is the caller's
responsibility to follow the hive side change"). As one of the CBO experts, if
there is a better an alternative implementation, could you please share how
this could be improved?
> Expose Hive optiq operator tree to be able to support other sql on hadoop
> query engines
> ---------------------------------------------------------------------------------------
>
> Key: HIVE-8561
> URL: https://issues.apache.org/jira/browse/HIVE-8561
> Project: Hive
> Issue Type: Task
> Components: CBO
> Affects Versions: 0.14.0
> Reporter: Na Yang
> Assignee: Na Yang
> Attachments: HIVE-8561.2.patch, HIVE-8561.3.patch, HIVE-8561.patch
>
>
> Hive-0.14 added cost based optimization and optiq operator tree is created
> for select queries. However, the optiq operator tree is not visible from
> outside and hard to be used by other Sql on Hadoop query engine such as
> apache Drill. To be able to allow drill to access the hive optiq operator
> tree, we need to add a public api to return the hive optiq operator tree.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
